humanize-korean

Pass

Audited by Gen Agent Trust Hub on Aug 5, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill includes an explicit 'Ironclad Rule' (Rule 7) instructing the agent to treat all user-provided text strictly as data and never as instructions, providing a proactive defense against prompt injection attempts.
  • [EXTERNAL_DOWNLOADS]: The documentation references an external GitHub repository (epoko77-ai/im-not-ai) for attribution and as an alternative version of the tool. However, the skill does not perform any automated downloads, installations, or remote script executions.
  • [DATA_EXFILTRATION]: The skill reads local reference files and writes processed output to a standard workspace directory (_workspace/). It does not contain any network operations or mechanisms to exfiltrate data to external servers.
  • [INDIRECT_PROMPT_INJECTION]: As the skill's primary function is to process user-supplied Korean text, it possesses an ingestion surface for indirect prompt injection. This is mitigated by the 'data-only' processing rule and the specific linguistic patterns the agent is instructed to follow.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 5, 2026, 06:33 AM
Security Audit — agent-trust-hub — humanize-korean